Jun 01, 2018
Thermal management is required to achieve optimal power electronic system performance and reliability. While in operation, power semiconductor devices generate a lot of waste heat as a result of conductive and switching losses. This heat has to be dissipated from the semiconductor junction to the semiconductor package and ultimately to the ambient environment to prevent thermal runaway. Of course, semiconductor technology keeps on improving and losses will be further reduced over the coming decades, but the size of the devices will shrink at an even faster pace causing an increase in heat flux density. As a consequence, design engineers will have to tackle these thermal issues and will need to identify new approaches to cool down their power devices.
Heat flux density
In theory, a large amount of heat could be removed from a hot device if the area and volume available for heat dissipation are unlimited. Unfortunately this is not likely to be the case in reality and particularly not with power semiconductors, high brightness LED and laser diodes which are assembled in small packages and integrated in systems with very limited space for cooling the package. Therefore, as we speak about the flow of heat, we should always consider it in per unit of area. For power semiconductor devices, heat flux density is typically in the range of 100W per cm².
Laser diode bars generate even more heat. They transform a given electrical input power into useful optical power. These bars are very small (typically 10mm x 1mm) and they have a low to medium power conversion efficiency. This inevitably means a large amount of heat is concentrated in a very small area. Today, heat flux density can typically reach up to 1000W per cm² and one can expect even higher heat flux densities for future product generations.
Heat transfer fundamentals
Transfer of thermal energy occurs only through conduction, convection, radiation or a combination of these three. Heat radiation is the emission of electromagnetic radiation from the surface of an object as a result of the random movements of its atoms and molecules due to the object’s temperature. Even though heat radiation can have a significant influence in many applications, its contribution to the overall heat dissipation in the case of power electronics can usually be neglected. Predominant mechanisms are heat conduction from semiconductor junction to case and heat convection from case to ambient.
Heat conduction from junction to case
Heat conduction is the transfer of heat through a medium from a region of higher temperature – in our case a small semiconductor device - to a region of lower temperature – usually a heat sink with a large surface or the bottom of the case which is exposed to a cooling medium. The medium can be a printed circuit board or any other kind of substrate on which the semiconductor device is attached. The temperature difference between the heat source and the bottom of the case is driving the heat flow but the substrate can provide resistance to the heat flow depending on its material grade, thickness and surface in contact with the coolant.
The material’s ability to conduct heat is described by its thermal conductivity. This is an intrinsic property of a material. This means that thermal conductivity is independent of material size, shape and orientation (but slightly decreases as temperature increases) for homogeneous materials. Metals like copper have a very good thermal conductivity of up to 400W/mK depending on copper purity. Copper is very well suited and widely used as lead frame or base plate for the packaging of semiconductors. However, semiconductors have to be electrically isolated from ground to comply with safety regulations. Unfortunately, materials which provide a good electrical isolation generally have a much lower thermal conductivity than copper. For instance, glass reinforced epoxy resins which are used in printed circuit boards have a thermal conductivity below 1W/mK. It can be increased by an order of magnitude when epoxy resins are filled with ceramic particles. The best thermal conductivity can be achieved by ceramics which are used in Direct Bond Copper (DBC) and Active Metal Brazed (AMB) substrates. Their thermal conductivity spans from 24W/mK for Al2O3 up to 180W/mK for AlN.
In addition, when these insulating layers have to be thick to fulfill the application’s specific requirements regarding electrical isolation, such layers provide a significant resistance to the heat flow. They act as thermal barriers in the path from the hot source to the bottom of the case. When heat encounters such a thermal barrier it starts looking for a less resistant way to escape from the high temperature region before being forced to go through the thermal barriers. This physical behavior is called “heat spreading” and plays an important role in power electronics.
Heat transfer from case to ambientUltimately, heat is spread and conducted to the bottom of the case – usually a heat sink: the larger the contact surface between the bottom of the case and the coolant is, the more heat is exposed to and can be exchanged with the coolant by convection. Heat sinks are available with different shapes and fins to increase the contact surface as much as possible. Heat transfer to the coolant then happens by convection. It not only depends on the temperature difference between the large contact surface and the coolant, it is also driven by the movement of the cooling fluids. We speak about “free convection” when fluid motion is caused by buoyancy forces that result from the density variations due to variations of temperature in the fluid. As opposed to free convection, “forced convection” occurs when the fluid is forced to flow over the hot surface by an external source such as fans or pumps. In both cases, the flow can be internal or external. Internal flow occurs when a fluid is enclosed by a solid boundary such when flowing through a pipe or channels. An external flow occurs when a fluid extends indefinitely without encountering a solid surface. Finally, we can distinguish between “laminar flow” and “turbulent flow”. The latter is characterized by chaotic changes in pressure and flow velocity. It is in contrast to a laminar flow which occurs when a fluid flows in parallel layers, with no disruption between those layers. Obviously the best heat dissipation is achieved by forced convection and with turbulent flow inside a structure.
The “heat transfer coefficient” is the ratio between the “heat flux” and the “thermodynamic driving force” for the flow of heat, in our case the temperature difference. Typical values for heat transfer coefficient vary from very few up to many thousands W/m²K. For a given fluid, forced convection is always better than free convection but higher costs are associated with the required external source for fluid movement. Liquid cooling outperforms air cooling by having heat transfer coefficients several orders of magnitude higher. However, liquid cooling comes with its own risks and potential problems, such as leakage, corrosion, and condensation. Therefore, liquid cooling is reserved for applications involving power densities that are too high for safe dissipation by air cooling. But liquid cooling has to be preferred to air cooling for applications where compact and light weight solutions are required. Examples of liquid cooling solutions include cold plates and tubes, jet impingement, macro and micro channel coolers.
While free convection in the air is easy to design and not expensive, it is a very popular cooling technology but it usually does not provide sufficient cooling performance for most of the power electronics applications. Forced air convection is the preferred cooling technology in industrial applications and liquid cooling is mandatory when designers are facing space constraints. Another very efficient heat transfer can also be achieved during a phase change of the cooling medium. In a two phase system, heat is transferred to the cooling medium as it changes its phase from fluid to gas. The cooling medium is then transported to an air cooled condenser where it changes its phase back again from gas to fluid. In such two phase systems, the transportation of the cooling medium is driven by a pump, gravity or capillary forces.
Find the right balance between heat conduction and heat convection
As you design your cooling system, you have to consider both the heat conduction from junction to case and the heat transfer from case to ambient. The whole thermal path from junction to ambient has to be taken into account. Generally speaking, for any given stack of material, the heat spreading effect is reduced and the contribution of heat conduction (from junction to case) to the whole thermal path (from junction to ambient) is increased as the heat transfer (from case to ambient) is improved. However, optimizing the heat transfer to the coolant may be very expensive and inefficient if you have very poor heat conduction from junction to case. And very expensive materials can provide very good heat conduction but don’t help that much if you have a low cost but poor thermal interface with the coolant. It is not an easy task to find the right balance between heat conduction and heat convection as the supply chain from the chip to the system is broken down in multiple subunits. In the last few years, design changes are happening rapidly as more and more compact modules with integrated cooling circuits are launched on the market to guarantee the required thermal performance so that the end user won’t experience any surprise or disappointment in product performance.
Do you have any design question or require some assistance for the selection of a suitable substrate or cooler for your application? Rogers PES’ experts are available to help. Please contact us today.